JP2010271798A - Distributed data management device, method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To efficiently read the same type of data such as a specific data source or a data type in a distributed data management system using a DHT (Distributed Hash Table). <P>SOLUTION: In each node of a distributed data management system using the DHT, during the writing of data, the writing is performed by using a value to be obtained from the value of data as a key, and furthermore, the writing of data is also performed by using a value to be obtained from metadata (the identification information of a data source and the identification information of a data type or the like) as a key, and during the reading of data, the reading is performed by designating the data source or the data type as necessary, so that access is performed only to the node or node group in which the necessary data are stored. Also, by storing data by using the value to be obtained from the metadata as key, it is possible to simultaneously create a replica. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a distributed data management apparatus, method, and program, and more particularly, to a distributed data management apparatus, method, and program for distributing and storing and managing data in a plurality of nodes that communicate via a network.

  In recent years, with the increase in the amount of digital data, the scalability of data management systems has become important, especially in terms of growth capability, peer-to-peer (P2P) type distributed data with excellent scale-out characteristics. Management systems are attracting attention.

  The P2P data management system is a well-known technology and can be roughly divided into two types: structured P2P (structured P2P) and unstructured P2P. Research and development of the structured P2P method using the (DHT: Distributed Hash Table) technology has become active.

  In these structured P2P data management systems using DHT, generally, the data itself is regarded as a bit string, and a fixed-length key value is generated from the data using a hash function or the like, and the data is based on the key value. A method for determining a node to be stored is adopted. The correspondence between the key value and the data storage node is managed in the routing table, and is updated as appropriate according to the participation and withdrawal of each node from the system. By using the hash function, it can be expected that the data is distributed evenly to each node constituting the system without the write load being concentrated on a specific node. Furthermore, in a structured P2P data management system using DHT, generally, each node does not need to know all the correspondences between key values and nodes. The node that has received the read or write request including the key refers to the routing table, and if the key value is within the management target range of the own node, process it in the own node, otherwise, the key. It has a mechanism to forward the request to a node corresponding to the value or a node that is highly likely to know the node corresponding to the key value. Thereby, even in an environment where participation and withdrawal of a node are frequently performed, it is possible to deliver a request to an appropriate node while suppressing the update frequency of the routing table (for example, Non-Patent Document 1, 2).

"A Survey and Comparison of Peer-to-Peer Overlay Network Schemes", Eng Keong Lua, Jon Crowcroft, Marcelo Pias, Ravi Sharma and Steven Lim, IEEE COMMUNICATIONS SURVEY AND TUTORIAL, MARCH 2004. "WIDE 2002 Research Report Part 17 Integrated Distributed Environment with Overlay Network"

  However, if the data itself is simply regarded as a bit string and the data is stored using the hash value as a key, when reading the data, a method other than specifying the key value and reading the data Does not increase. For example, in a system that manages the output data of a large number of sensors, when it is desired to read the output data of only a specific sensor, the data of the specific sensor exists in different nodes in the system according to the hash value of the value. Whether or not the data of the specific sensor is possessed is not known unless the individual nodes are accessed.

  That is, in order to read only the data of a specific sensor, it is necessary to access all the nodes in the system, and degradation of search efficiency becomes a problem.

  Also, when processing data, processing programs often only handle specific types of data. However, just by storing data using a hash value that is regarded as a bit string as a key, as described in the previous section, it is not possible to specify a node having the corresponding type of data in advance. Access is necessary, and degradation of search efficiency becomes a problem.

  In addition, if data is stored only by using a hash value as a key, if a failure occurs in the storage node, access to the data becomes impossible. Conventionally, in order to solve this, the hash value is determined in advance. A method has been adopted in which a value obtained by performing the calculated operation is used as a key and a replica (replica) is stored in a node corresponding to the key.

  The present invention has been made in view of the above points, and in a distributed data management system using DHT, a distributed data management apparatus capable of efficiently reading the same type of data such as a specific data source and data type. And a method and a program.

  FIG. 1 is a principle configuration diagram of the present invention.

The present invention (Claim 1) is a distributed data management device (node) of a distributed data management system composed of a plurality of nodes,
The managed data includes metadata,
Communication / interface means 11 for exchanging data with the outside of the node;
Key generation means 14 for generating a key from an arbitrary bit string by a predetermined method;
A database 16 for storing key and data pairs;
Database management means 13 for storing a key-data pair in the database 16 and retrieving data corresponding to the designated key from the database;
Routing means 15 for routing to other nodes corresponding to the specified key using a routing table 17 based on the distributed hash table technology (DHT);
A request processing unit 12 that analyzes a write request and a read request acquired via the communication / interface unit 11 and controls the communication / interface unit 11, the key generation unit 14, the database management unit 13, and the routing unit 15; ,
The request processing means 12
When a write request is acquired from an external node via the communication / interface unit 11, the key generation unit 14 generates a key from the data included in the request, and the routing unit 15 generates a key based on the key. Means for instructing routing to a node to be written, and sending a write request comprising a key and data to the node to be written via the communication / interface means 11;
If it is determined that the key is included in the write request, the key is within the management target range of the own node, and the request is not a replica (replication) write request, the database management means The key and the data included in the request are instructed to be written in the database, the key generation unit 14 is instructed to generate a key using all or part of the metadata of the data as a byte string, and the routing unit 15 For the replica writing target node based on the key, and the communication / interface means 11 uses the key and the data as a write request as a replica write request. And the routing means 15 associates the key of the data and the key of the replica with the route. And means for updating the I ring table 17,
When it is determined that the write request includes a key, the key is within the management target range of the own node, and the request is a replica write request, the database management unit 13 is Means for instructing the database 16 to write the data included in the request;
When a data read request is acquired and it is determined that the key of the request is outside the management target range of the own node, the routing unit 15 is instructed to route to the read target node based on the key, and the communication / interface unit Means for transferring the read request to the read target node via 11 and transferring a response returned from the read target node to the request source;
If the key of the data read request is within the management target range of the own node, the database management unit 13 is instructed to read the corresponding data from the database 16 based on the key, and the communication / interface unit 11 includes the data. Means for sending a response to the forwarding node;
A distributed data management device comprising:

  In the present invention (Claim 2), when data is written into the database in the database management means, the value obtained from the metadata including the identification information of the data source or the identification information of the data type is obtained. Write as a key.

  According to the present invention (Claim 3), when the database management means reads from the database, the data source or data type is designated and the reading is performed.

  FIG. 2 is a diagram for explaining the principle of the present invention.

The present invention (Claim 4) is a distributed data management method in each node of a distributed data management system composed of a plurality of nodes,
Managed data includes metadata,
When a write request is received from an external node (step 1), a key is generated from the data included in the request (step 2), and routing to the write target node is performed based on the key (step 3). A write request including the key and the data is transmitted to the write target node (step 4), and a response from the write target node is transmitted to the request source (step 5).
When the key is included in the write request, the key is within the management target range of the own node (step 6, Yes), and the request is not a replica write request (step 7, No), the Write the key and the data included in the request to the database (step 8), generate a key using all or part of the metadata of the data as a byte string (step 9), and write the replica to the node based on the key (Step 10), the write request comprising the key and the data is transferred to the replica write target node as a replica write request (step 11), and the response from the write target node is transmitted to the source node. (Step 12), a write request including the key and the data is transmitted to the write target node, It associates the data key and the replica of the key to update the routing table (step 13),
When a key is included in the write request, the key is within the management target range of the own node (step 6, Yes), and the request is a replica write request (step 7, Yes), Write data included in the request to the database (step 14), send a response to the source node (step 22),
When the key is included in the write request and the key is not within the management target range of the own node (No in step 6), routing to the write target node is performed (step 19), and the write target node is reached. The request is transferred (step 20), and the response from the write target node is transmitted to the transmission source node (step 21).
When the data read request is acquired, if the key of the request is out of the management target range of the own node, routing to the read target node is performed based on the key (step 15), and the read request is transmitted to the read target node. (Step 16), the response returned from the read target node is transferred to the request source,
If the key of the data read request is within the management target range of the own node, the corresponding data is read from the database based on the key (step 17), and a response including the data is transmitted to the request source node (step 18).

  Further, according to the present invention (Claim 5), when data is written to the database, the data source identification information or the value obtained from the metadata including the data type identification information is used as a key. .

  According to the present invention (Claim 6), when reading from the database, the data source or data type is designated and the reading is performed.

  The present invention (Claim 7) is a distributed data management program for causing a computer to function as each means constituting the distributed data management apparatus according to any one of Claims 1 to 3.

  As described above, according to the present invention, in each node of the distributed data management system using DHT, at the time of data writing, writing is performed using the value obtained from the data value as a key, and further, metadata (data source identification) is performed. Information, data type identification information, etc.) is used as a key to write data. When reading data, the data source and data type can be specified as necessary to read data. By accessing only the stored node or group of nodes, the search efficiency is further improved.

  Furthermore, in general, distributed data management systems aim to improve reliability and the like by storing replicas of normal data in a plurality of different nodes, but in conventional distributed data management systems using DHT, replica placement Employs a method of storing a replica in a node corresponding to the key using a value obtained by performing a predetermined calculation on the hash value (original key) derived from the data. However, in the distributed data management system of the present invention, it is possible to simultaneously create a replica by storing data using a value obtained from the above-described metadata (data source identification information, data type identification information, etc.) as a key. Is.

  As described above, according to the present invention, in the distributed data management system, it is possible to improve the efficiency of narrowing down during data reading and improve the reliability performance while maintaining the dispersibility of data writing. Conventionally, a method has been adopted in which a value obtained by performing a predetermined operation on the hash value is used as a key, and a replica is stored in a node corresponding to the key. However, in the present invention, instead of preparing a replica by this method, when storing data using a value obtained from the above-mentioned metadata (data source identification information, data type identification information, etc.) as a key, the replica is used. By configuring the routing table so that it can be used as well, the availability of the data can be increased.

It is a principle block diagram of this invention. It is a figure for demonstrating the principle of this invention. 1 is a configuration diagram of a distributed data management system to which the present invention is applied. FIG. It is a block diagram of the node in one embodiment of this invention. It is the data (the 1) of the temperature sensor in one embodiment of this invention. It is the data (the 2) of the temperature sensor in one embodiment of this invention. It is a sequence chart of the writing method in one embodiment of the present invention. It is a sequence chart of a reading method when a key generated from data is specified. 6 is a sequence chart (specifying a key generated from metadata) of a reading method according to an embodiment of the present invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 3 shows the configuration of a distributed data management system to which the present invention is applied, and FIG. 4 shows the configuration of a node in an embodiment of the present invention.

  In the present invention, the node shown in FIG. 4 is operated on one PC (personal computer), and these PCs are connected via an IP (Internet Protocol) network, so that the data management system shown in FIG. Constitute. Here, the PC may be a single hardware, or may be a virtual machine using a virtualization technique. Virtual machines are already well-known technologies. For details, see `` http://en.wikipedia.org/wiki/%E4%BB%AE%E6%83%B3%E6%A9%9F% The detailed description is omitted here.

  The node 10 includes a communication / interface unit 11, a request processing unit 12, a database management unit 13, a key generation unit 14, a routing unit 15, a database 16, and a routing table 17. Among these, the database 16 and the routing table 17 are storage media such as a hard disk device.

  The communication / interface unit 11 exchanges requests and responses with the outside of the node.

  The request processing unit 12 determines whether the external request received by the communication / interface 11 is a write request, a read request, whether the request includes a key, and if the key is included, Whether or not the key is within the management target range of the own node or a write request, it is determined whether or not the request is a replica write request, and the communication / interface 11, database management unit 13, key generation unit 14, routing unit 15 is controlled.

  The database management unit 13 performs a process of writing a key / data pair in the database 16 by a write request and a process of reading data corresponding to the key specified by the read request from the database 16.

  The key generation unit 14 generates a key from an arbitrary bit string by a predetermined method.

  The routing unit 15 performs routing to another node corresponding to the designated key using the routing table 17 based on the distributed hash table technique (DHT).

To input and get data to the above system from outside the system,
(A) A method of temporarily joining the system as a node and leaving the system after inputting data; (b) The communication / interface unit 11 of the node 10 has a user interface function or a program interface function as a part of the function. , How to input and get data through this;
Shall be used.

  In the present invention, metadata is used as management target data. The metadata includes data identification information, data type identification information, and the like. In this embodiment, the ID (identifier) of the data source is used. As the data source ID, URN (reference document: IETF RFC 2141-URN Syntax: http://www.ietf.org/rfc/rfc2141.txt) or the like is used and is determined in advance outside the system.

  As a hash algorithm for generating a key used in the key generation unit 14, MD5 which is an existing hash function, SHA family (SHA-1, SHA-224, SHA-256, SHA-384, SHA-512), etc. Any one of them shall be used.

  This will be specifically described below.

  It is now assumed that temperature data is input to the system from two different temperature sensors. The first data is the data shown in FIG. 5 from the temperature sensor whose data source ID is 'urn: thermosensor: 9999', and the second data is from the temperature sensor whose data source ID is 'urn: thermosensor: 8888'. Assume that the data is shown in FIG.

  FIG. 7 is a sequence chart of the writing method according to the embodiment of the present invention.

  Hereinafter, the flow of storing the first data in the distributed data management system will be described with reference to FIG. In the following, each node 10 will be described with a node ID.

  Step 101) First, the request processing unit 12 receives a write request including the data of FIG. 5 via the communication / interface unit 11 of the arbitrary node 321.

  Step 102) The key generation unit 14 of the node 321 considers the data in FIG. 5 as a bit string obtained by combining the XML data and the data source ID, calculates a hash using SHA-1, and obtains the value “12345”. Suppose that This value is the key for this data.

  Step 103) Next, the routing unit 15 of the node 321 performs routing using the routing table 17 in order to obtain a node corresponding to the key “12345”, and derives that the node 997 is a corresponding node. To do.

  Step 104) Therefore, the node 321 transmits a write request including the key “12345” and the data shown in FIG. 5 to the node 987.

  Step 105) Upon receiving the write request, the request processing unit 12 of the node 987 indicates that it is a write request including the key “12345” and that the key is a management target of the own node via the routing unit 15. After confirming with the routing table 17, the data is stored in the database 16 of the own node in association with the key “12345” via the database management unit 13.

  Step 106) Next, the key generation unit 14 of the node 987 considers the data source ID (urn: thermosensor: 9999), which is metadata, as a bit string, calculates a hash using SHA-1, and calculates “98765”. Is obtained. This value is the key for this data.

  Step 107) Next, it is assumed that the routing unit 15 of the node 987 performs routing using the routing table 17 in order to obtain this key “98765” and derives that the node 111 is a corresponding node.

  Step 108) Accordingly, the node 987 transmits a replica write request including the key “98765” and the data shown in FIG.

  Step 109) Upon receiving the write request, the request processing unit 12 of the node 111 confirms that the request is a write request including the key “98765” and that the key is a management target of the own node via the routing unit 15. After confirming with the routing table 17, the data is stored in the database 16 of the own node in association with the key “98765”.

  Similarly to the above, the flow in which the second data (data source ID: urn: thermosensor: 8888) shown in FIG. 6 is stored in the distributed data management system will be described along the sequence chart of FIG.

  Step 101) First, the request processing unit 12 of an arbitrary node (for example, the node 456) receives a write request including the data of FIG. 6 via the communication / interface unit 11.

  Step 102) The key generation unit 14 of the node 456 considers the data in FIG. 6 as a bit string of data obtained by combining the XML data and the data source ID, calculates a hash using SHA-1, and obtains “13579”. Suppose you get a value. This value is the key for this data.

  Step 103) Next, the routing unit 15 of the node 456 performs routing using the routing table 17 in order to obtain a node corresponding to the key “13579”, and derives that the node 654 is a corresponding node. .

  Step 104) Therefore, the node 456 transmits a write request including the key “13579” and the data of FIG. 6 to the node 654.

  Step 105) The node 654 that has received the write request confirms by the routing table 17 that it is a write request including the key “13579” and that the key is a management target of its own node. The data is stored in the node database 16 in association with the key “13579”.

  Step 106) The key generation unit 14 of the node 654 considers the data source ID (urn: thermosensor: 8888), which is metadata, as a bit string, performs hash calculation using SH1-1, and obtains a value “54321”. Suppose. This value is the key for this data.

  Step 107) Next, the routing unit 15 of the node 654 performs routing using the routing table 17 in order to obtain a node corresponding to the acquired key “54321”, and derives that the node 222 is a corresponding node. To do.

  Step 108) Therefore, the node 654 transmits a replica write request including the key “54321” and the data of FIG. 5 to the node 222.

  Step 109) The node 222 that has received the replica write request confirms by the routing table 17 that it is a write request including the key “54321” and that “54321” is the management target of the own node. Then, the data is stored in the database 16 of the own node in association with the key “54321”.

  Hereinafter, when different data is input from the sensor 'urn: thermosensor: 9999', the data value is different, so the data storage node (the node specified by the key calculated from the XML data and the data source ID) is not necessarily required. Although not necessarily the same node, the storage node of the replica (the node specified by the key calculated from the data source ID) is always the node 111. Similarly, when another data is input from the sensor 'urn: thermosensor: 9999', the storage node of the replica is always the node 222.

  Using this property, when data input from a specific sensor is read out from the distributed data management system of the present invention (as shown in FIG. 8, when only storing based on the data key is performed, a read request is sent to all nodes). In the system of the present invention, as shown in FIG. 9, a metadata key, ie a key derived from a data source ID (eg urn: thermosensor: 9999) (eg “98765”) ) Can be used to read data only from the node where the corresponding data exists (for example, node ID: 111).

  In the present invention, the replica position is recorded in the routing table 17 when the replica is inserted by the metadata key. Since the routing table 17 is appropriately synchronized between nodes based on the DHT mechanism, when the corresponding storage destination node does not respond to the read request by the data key for some reason, the routing table 17 starts from the storage destination node corresponding to the metadata key. When data is read, and conversely, when a corresponding storage destination node does not respond to a read request by a metadata key for some reason, it is possible to read data from the storage destination node corresponding to the data key. . For example, when data reading is attempted using the key “12345” derived from the data in FIG. 5, but the corresponding node 987 does not respond due to a failure, the node that has received the read request sends a replica from the routing table 17. The key of “98765” is obtained, and information that the corresponding node is ID = 111 is obtained, a read request is transferred to the node 111, and data can be read.

  The operation of the components of the node 10 shown in FIG. 4 is constructed as a program and can be installed and executed on a computer used as a node (distributed data management apparatus), or distributed via a network. Is possible.

  Further, the constructed program can be stored in a portable storage medium such as a hard disk, a flexible disk, or a CD-ROM, and can be installed or distributed in a computer.

  The present invention is not limited to the above-described embodiment, and various modifications and applications can be made within the scope of the claims.

10 nodes (distributed data management device)
11 Communication / Interface Unit, Communication / Interface Unit 12 Request Processing Unit, Request Processing Unit 13 Database Management Unit, Database Management Unit 14 Key Generation Unit, Key Generation Unit 15 Routing Unit, Routing Unit 16 Database 17 Routing Table

Claims (7)

A distributed data management device (node) of a distributed data management system composed of a plurality of nodes,
The managed data includes metadata,
A communication / interface means for exchanging data with the outside of the node;
Key generating means for generating a key from an arbitrary bit string by a predetermined method;
A database that stores key and data pairs;
Database management means for storing a key and data pair in the database and retrieving data corresponding to the specified key from the database;
A routing means for routing to another node corresponding to the specified key using a routing table based on the distributed hash table technology;
A request processing unit that analyzes the write request and the read request acquired through the communication / interface unit, and controls the communication / interface unit, the key generation unit, the database management unit, and the routing unit;
The request processing means includes:
When a write request is acquired from an external node through communication / interface means, the key generation means is caused to generate a key from data included in the request, and the routing means is written based on the key. Means for instructing routing to a target node, and causing the write target node to transmit a write request comprising the key and the data via the communication / interface means;
When it is determined that the key is included in the write request, the key is within the management target range of the own node, and the request is not a replica write request, the database management means Instructing the key and the data included in the request to be written into the database, instructing the key generation unit to generate a key using all or part of the metadata of the data as a byte string, and the routing unit To the replica write target node based on the key, and to the communication / interface means, the write request comprising the key and the data as a replica write request, the replica write target And the routing means associates the key of the data with the key of the replica, and And means for updating the computing table,
When it is determined that the write request includes a key, the key is within the management target range of the own node, and the request is a replica write request, the database management unit Means for instructing to write the data included in the request to the database;
When a data read request is acquired and it is determined that the key of the request is outside the management target range of the node, the routing unit is instructed to route to the read target node based on the key, and the communication / interface Means for transferring the read request to the read target node via a means, and transferring a response returned from the read target node to the transmission source;
If the key of the data read request is within the management target range of the own node, the database management means is instructed to read the corresponding data from the database based on the key, and the data is sent to the communication / interface means. Means for sending a response to the requesting node;
A distributed data management device comprising: The database management means includes:
The distributed data management apparatus according to claim 1, wherein when data is written to the database, writing is performed using a value obtained from the metadata including data source identification information or data type identification information as a key. The database management means includes:
2. The distributed data management apparatus according to claim 1, wherein when reading from the database, reading is performed by designating a data source or a data type. A distributed data management method in each node of a distributed data management system composed of a plurality of nodes,
Managed data includes metadata,
When a write request is received from an external node, a key is generated from the data included in the request, routing to the write target node is performed based on the key, and the write request consisting of the key and the data is sent to the write target. Send the response from the write node to the requestor,
If the key is included in the write request, the key is within the management target range of the own node, and the request is not a replica write request, the key and the data included in the request are stored in the database. , Generate a key using all or part of the metadata of the data as a byte sequence, perform routing to the replica write target node based on the key, and send a write request comprising the key and the data to the replica As a write request, transfer to the replica write target node, send a response from the write target node to the source node, send a write request consisting of the key and the data to the write target node, and And update the routing table in association with the replica key,
When the key is included in the write request, the key is within the management target range of the own node, and the request is a replica write request, the data included in the request is written to the database. Send a response to the source node,
If the write request includes a key and the key is not within the management target range of the own node, routing to the write target node is performed, the request is transferred to the write target node, and the write target node Send the response from to the source node,
When the data read request is acquired, if the key of the request is out of the management target range of the own node, routing to the read target node is performed based on the key, and the read request is transferred to the read target node. The response returned from the read target node is transferred to the request source,
If the key of the data read request is within the management target range of the own node, the corresponding data is read from the database based on the key, and a response including the data is transmitted to the request source node.
And a distributed data management method. 5. The distributed data management method according to claim 4, wherein when data is written to the database, writing is performed using a value obtained from the metadata including identification information of a data source or identification information of a data type as a key. The distributed data management method according to claim 4, wherein when reading from the database, reading is performed by designating a data source or a data type.   The distributed data management program for functioning a computer as each means which comprises the distributed data management apparatus of any one of Claims 1 thru | or 3.

Images